Process-Guided Concept Bottleneck Model
This work addresses the problem of improving explainability and trustworthiness in AI for scientific applications where supervision is sparse, representing an incremental advancement over standard CBMs.
The paper tackles the limitations of Concept Bottleneck Models (CBMs) in scientific domains by proposing a Process-Guided CBM (PG-CBM) that incorporates domain-specific causal mechanisms, using biomass estimation from Earth Observation data as a case study, and shows it reduces error and bias compared to benchmarks while improving interpretability.
Concept Bottleneck Models (CBMs) improve the explainability of black-box Deep Learning (DL) by introducing intermediate semantic concepts. However, standard CBMs often overlook domain-specific relationships and causal mechanisms, and their dependence on complete concept labels limits applicability in scientific domains where supervision is sparse but processes are well defined. To address this, we propose the Process-Guided Concept Bottleneck Model (PG-CBM), an extension of CBMs which constrains learning to follow domain-defined causal mechanisms through biophysically meaningful intermediate concepts. Using above ground biomass density estimation from Earth Observation data as a case study, we show that PG-CBM reduces error and bias compared to multiple benchmarks, whilst leveraging multi-source heterogeneous training data and producing interpretable intermediate outputs. Beyond improved accuracy, PG-CBM enhances transparency, enables detection of spurious learning, and provides scientific insights, representing a step toward more trustworthy AI systems in scientific applications.